Abstract
It has been suggested that aberrant inferences (circular inferences) implemented in brain networks, generating a system that overweights even weak sensory evidence could underlie the positive dimension of schizophrenia (Jardri and Denève, 2013). This hypothesis was validated by a probabilistic reasoning task (in prep.). Here we validate, with a pilot study in healthy subjects, a paradigm that allows for experimentally testing the respective impacts of sensory evidence and priors on perception in schizophrenia. Necker Cube is an ambiguous figure, known to induce bistability. Such figures were continuously presented to 50 participants during 15 consecutive runs. We manipulated sensory evidence by adding shades to the stimuli (3 last runs) and prior expectations by giving different instructions to 3 different groups (15 – 15 – 20 participants), concerning the presence of an implicit preference. Participants' responses were discretely and pseudo-regularly collected (Mamassian and Goutcher, 2005). After confirming the existence of an implicit prior (p< 0.001), we showed that manipulation of this prior had significant opposite effects (p=0.009), either by exacerbating or cancelling the intrinsic bias of the system. The effect of sensory evidence was even stronger (p< 0.001), and induced a significant bias corresponding to the direction of the cue. This effect was so strong that it overcame the impact of the instructions when cues and prior manipulations were combined. We also found that the behavior could be well fitted by Bayesian models ("simple" Bayes, model with Markovian statistics) with low statistical dependencies between successive time steps. This study will be used as a reference, in order to study patients with psychotic symptoms and test our initial claim.
Meeting abstract presented at VSS 2016